Folks,
only now was I able to do regression tests regarding my proposed fix(es)
of the misplaced-note-head-bug (issue 5303):
The bug is caused by the integer-cast
lastpos = int (p);
in lily/stem.cc:629.
Replacing it by either
i) lastpos = p; // use float value
or
ii) lastpos = int (round (p)) // correctly round to integer
resolves the bug. Neither variant introduces regressions (differences
only in in test-output-distance and rest-dot-position, the latter being,
I think, unrelated - the seemingly random position changes of dots to
multi-voice simultaneous rests).
Questions:
a) How to decide which fix is "better"? (My guess that using floats
might pose the danger of rounding errors adding up until something bad
happens - probably only for huge chords comprising hundreds of notes,
but I'd tend to favor the solution with rounding.)
b) Would anybody be willing to shepherd a patch through the review process?
c) Does the situation warrant inclusion of a new regression test? (The
example in the bug report on
https://sourceforge.net/p/testlilyissues/issues/5303/ would be suitable
I think.)
Best
Lukas
diff --git a/lily/stem.cc b/lily/stem.cc
index 37aa40c250..2a0f6fef84 100644
--- a/lily/stem.cc
+++ b/lily/stem.cc
@@ -626,7 +626,7 @@ Stem::calc_positioning_done (SCM smob)
else
parity = true;
- lastpos = int (p);
+ lastpos = p;
}
return SCM_BOOL_T;
diff --git a/lily/stem.cc b/lily/stem.cc
index 37aa40c250..34bf557fca 100644
--- a/lily/stem.cc
+++ b/lily/stem.cc
@@ -626,7 +626,7 @@ Stem::calc_positioning_done (SCM smob)
else
parity = true;
- lastpos = int (p);
+ lastpos = int (round (p));
}
return SCM_BOOL_T;
_______________________________________________
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel
